Expand description
Weight vector operations for meta-learning and target network updates.
Provides Reptile-style meta updates, Polyak (exponential moving average) updates for SAC/TD3 target networks, and weight vector averaging.
Structs§
- Polyak
Update - Polyak (EMA) target network update strategy.
- Reptile
Update - Reptile meta-learning update strategy.
Traits§
- Weight
Update - Trait for weight vector update strategies.
Functions§
- average_
weight_ vectors - Average N weight vectors element-wise:
result[i] = mean(vectors[j][i] for all j) - polyak_
update - Exponential moving average (Polyak update):
target[i] = tau * source[i] + (1 - tau) * target[i] - reptile_
update - Reptile weight update:
params += lr * (task_params - params)